mlsanityData to 7 Oct 2026

Is Chain-of-Thought Really Not Explainability? Chain-of-Thought Can Be Faithful without Hint Verbalization

arXiv 2512.23032 ↗First posted 28 Dec 2025cs.CL cs.AI cs.LGcited by 15

On the map: LLM reasoning → Prompting and chain-of-thought reasoning

Abstract (arXiv)

Recent work, using the Biasing Features metric, labels a CoT as unfaithful if it omits a prompt-injected hint that affected the prediction. We argue this metric adopts a narrow notion of faithfulness and confuses unfaithfulness with incompleteness, the lossy compression needed to turn distributed transformer computation into a linear natural language narrative. On multi-hop reasoning tasks with instruct-tuned and reasoning models, many CoTs flagged as unfaithful by Biasing Features are judged faithful by other metrics, exceeding 50% in some models. With a new faithful@k metric, we show that larger inference-time budgets greatly increase hint verbalization (up to 90% in some settings), suggesting much apparent unfaithfulness is due to tight token limits. Using Causal Mediation Analysis, we further show that even non-verbalized hints can causally mediate prediction changes through the CoT. We therefore caution against relying solely on hint-based evaluations and advocate a broader interpretability toolkit, including causal mediation and corruption-based metrics. We do not claim all CoTs are faithful, only that the absence of hint words alone does not prove unfaithfulness.

Key facts. Cited by 15 papers in the map (data through 7 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 6 of its 45 references that are in the map. First posted on arXiv 28 Dec 2025.

Builds on

6 of 45 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Chain-of-Thought Reasoning In The Wild Is Not Always Faithful11 Mar

Other references · 5

  1. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,710
  2. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  3. 2022Demonstrate-Search-Predict: Composing retrieval and language models for knowledge-intensive NLPcited by 316
  4. 2023AI Control: Improving Safety Despite Intentional Subversioncited by 207
  5. 2025Agentic Misalignment: How LLMs Could Be Insider Threatscited by 136

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 15.

15 papers.

Sort:

By area:

  1. 2026RAGEN-2: Reasoning Collapse in Agentic RLcited by 16
  2. 2026Counterfactual Simulation Training for Chain-of-Thought Faithfulnesscited by 15
  3. 2026Lie to Me: How Faithful Is Chain-of-Thought Reasoning in Reasoning Models?cited by 8
  4. 2026Measuring Faithfulness Depends on How You Measure: Classifier Sensitivity in LLM Chain-of-Thought Evaluationcited by 4
  5. 2026Faithfulness Metrics Don't Measure Faithfulness: A Meta-Evaluation with Ground Truthcited by 3
  6. 2026When Shallow Wins: Silent Failures and the Depth-Accuracy Paradox in Latent Reasoningcited by 2
  7. 2026Investigating the Interplay between Contextual and Parametric Chain-of-Thought Faithfulness under Optimizationcited by 2
  8. 2026Measuring AI Reasoning: A Guide for Researcherscited by 1
  9. 2026Local Causal Attribution of Chain-of-Thought Reasoningcited by 1
  10. 2026Mechanistic Interpretability of Chain-of-Thought Reasoning via Sequential Activation Patchingcited by 1
  11. 2026LLM Reasoning Is Latent, Not the Chain of Thoughtcited by 0
  12. 2026Two Regimes of Chain-of-Thought Unfaithfulness: Metric-Based Detection Fails Where Models Are Wrongcited by 0
3 more
  1. 2026Chain-of-Thought Faithfulness of Reasoning Models Varies with Where and How Preference Cues Are Deliveredcited by 0
  2. 2026Untangling the Mechanisms of Misleading Context in Medical Question Answeringcited by 0
  3. 2026Making LLMs Say What They Think: Measuring and Improving CoT-Interpretability Alignmentcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting
  2. 2026Faithfulness as Information Flow: Evaluating and Training Faithful Chain-of-Thought Reasoning
  3. 2025FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning
  4. 2026Bypassing the Rationale: Causal Auditing of Implicit Reasoning in Language Models
  5. 2023Measuring Faithfulness in Chain-of-Thought Reasoning
  6. 2026Mechanistic Evidence for Faithfulness Decay in Chain-of-Thought Reasoning
  7. 2026FACT-E: Causality-Inspired Evaluation for Trustworthy Chain-of-Thought Reasoning
  8. 2025How does Chain of Thought Think? Mechanistic Interpretability of Chain-of-Thought Reasoning with Sparse Autoencoding